基于FCOS算法的乳腺癌影像病灶检测

Detection of Breast Cancer Image Focus Based on FCOS Algorithm

针对目前乳腺癌医学影像病灶检测算法中存在的筛查效率及准确率欠佳的问题,提出将一种改进的目标检测算法FCOS算法应用于乳腺癌超声影像病灶检测。首先,在将训练数据输入网络之前进行数据增强以减少过拟合。其次,将图片和对应的目标框输入图像预处理模块,再把结果传入FCOS中进行训练。最后直接将图片传入FCOS中即可进行使用。结果表明,FCOS算法的平均精度均值为86.23%,优于以往研究。

Aiming at the problem of low efficiency and accuracy in breast cancer medical image detection algorithm,this paper proposes an improved target detection algorithm FCOS algorithm applied to breast cancer ultrasonic image focus detection.First,the data is enhanced before the training data is input into the network to reduce over fitting.Secondly,the image and the corresponding target box are input into the image preprocessing module,and then the results are transferred to FCOS for training.Finally,the image can be directly transferred to FCOS for use.The results show that the average accuracy of FCOS is 86.23%,which is better than the previous research.